A probabilistic approach for solving Poisson equations in computer vision problems

摘要

Describes the probabilistic approach to solve Poisson equations and show this method may be used to solve computer vision problems. The authors also give a complexity analysis of this method and compare it method with BEM, FEM and FDM in terms of storage and time complexity. Some examples are given and show that the authors method is better than BEM, FEM and FDM in certain aspects. Finally, the authors use the probabilistic approach for solving several computer vision problems, such as shape from shading, surface interpolation and brightness based stereo vision, the experimental results show that the probabilistic approach is very effective.